Data_Revised Tunable Q-factor wavelet transform for EEG-based epileptic seizure detection

Citation Author(s):
Submitted by:
Zhen Liu
Last updated:
Fri, 09/16/2022 - 03:52
0 ratings - Please login to submit your rating.


This is a code to correlate and process a dataset of epilepsy EEG signals。


Description of the Epilepsy_DATA

It mainly includes the original BONN epilepsy EEG data set, as well as the EEG signals after the revised tunable Q-factor wavelet transform decomposition and reconstruction of the data set.






The dataset consists of five sub-data sets, each sub-data set contains 100 experimental samples, each sample is a single-channel EEG with a sampling frequency of 173.61 Hz and a duration of 23.6 seconds;

These fragments were manually sheared from long-range multichannel EEG. During the clipping process, some possible interferences are removed at the same time, including muscle artifact and eye movement artifact.

Subdatasets A and B are data collected from 5 healthy individuals, serving as a control group. Scalp electrode distribution is an international 10-20 system. In A, the EEG signal is collected when the subject's eyes are opened, and in B, the EEG signal is collected when the subject's eyes are closed.

Subset C/D/E is intracranial EEG, collected from 5 preoperatively diagnosed patients. These patients have had their epilepsy fully controllable by partial hippocampal resection. The resection area has been clinically verified as epileptogenic foci. Subset D contains EEGs collected from the epilepsy foci, and subset C contains EEGs collected from the contralateral side of the epileptic foci. Collection of C and D and interval between seizures. Subset E included ictal EEG. Subsets C and D were collected from deep electrodes. At the same time as the placement of the deep electrodes, some strip electrodes were also placed in the lateral and basal regions of the neocortical layer, and subset E contains the ictal EEG acquired by all the intracranial electrodes.






The above five datasets are in *.csv format, corresponding to the data obtained by the revised tunable Q-factor wavelet transform of the BONN dataset respectively. AB-CD-E means that AB is a type of data in BONN, CD is a type of data, and E is a type of data.


In the file “Data_Feature_extraction”, there are five data sets (AB-CD-E, A-C-E, A-D-E, B-C-E, B-D-E) through six time-frequency representations (CWT, DWT, EMD, FT, TQWT, RTQWT). feature data. These data are trained using the five machine learning methods (DT, LDA, NB, SVM, KNN) of the classification learners toolbox in matlab, and the resulting picture is the "Draw_Bonn_acc.fig" file in the file.